NVIDIA account plan
Accelerated computing technology
At a glance
NVIDIA is a data center scale AI infrastructure company with fiscal 2026 revenue of $215,938 million and about 42,000 employees in 38 countries.
NVIDIA, headquartered in Santa Clara, California, is a fabless platform company selling GPUs, CPUs (Grace, Vera), DPUs, NVLink, InfiniBand and Ethernet networking, rack-scale systems and software such as CUDA and NVIDIA AI Enterprise. Its customers are cloud providers, AI model makers, neoclouds, enterprises, the public sector, gamers and carmakers, served directly and through OEMs, ODMs and system integrators. In fiscal 2026 (year to January 25, 2026) revenue was $215,938 million (+65%), operating income $130,387 million and net income $120,067 million. It has about 42,000 employees in 38 countries, 31,000 of them in R&D. Co-founder Jensen Huang has been President and CEO since 1993.
Business units
Data Center
about 90% of fiscal 2026 revenue ($193,737 million, computed)Accelerated compute and networking, usually sold as rack-scale systems, plus software, to cloud providers, AI model makers, enterprises and the public sector. Q2 FY27 revenue was $89,023 million (+117%), split into Hyperscale ($48,710 million) and AI Clouds, Industrial, & Enterprise ($40,313 million). Blackwell is the vast majority of revenue, and Vera Rubin began production shipments in Q3 FY27.
Gaming
about 7% of fiscal 2026 revenue ($16,042 million, computed)GeForce RTX GPUs for desktop and laptop PCs, the GeForce NOW cloud gaming service and console SoCs; revenue grew 41% in fiscal 2026 on Blackwell demand. NVIDIA expects supply constraints to be a headwind to Gaming in fiscal 2027.
Professional Visualization
about 1.5% of fiscal 2026 revenue ($3,191 million, computed)RTX PRO workstation GPUs and software for design, engineering and content creation, plus Omniverse; revenue grew 70% on Blackwell demand and the launch of DGX Spark.
Automotive
about 1% of fiscal 2026 revenue ($2,349 million, computed)The DRIVE Hyperion platform (in-car computers, DRIVE OS, sensors and autonomous driving software) and data-center GPUs to train driving models, sold to carmakers, robotaxis and tier-1 suppliers; revenue grew 39%.
Strategic priorities
- One-year product cadence
- Move from Blackwell and Blackwell Ultra to Vera Rubin, which began production shipments in Q3 FY27 and is expected to be about 20% of Data Center revenue in Q3.
- Full-stack AI factories
- Sell the whole AI factory (Vera CPU, Rubin GPU, NVLink, InfiniBand, Ethernet and the Groq LPU), raising revenue opportunity per gigawatt from about $25 billion with Blackwell to $40 billion with Vera Rubin.
- Grow beyond hyperscalers
- Sovereigns, regional neoclouds, enterprises, edge and air-gapped data centers are expected to represent roughly half of the data center business; sovereign AI more than tripled year on year.
- CPUs and inference
- Grace CPU revenue exceeded $5 billion over twelve months and CPU revenue is expected to more than double in fiscal 2028; Groq 3 LPX racks extend inference to very low latency.
- Protect the CUDA install base
- Over 7.5 million developers use CUDA and NVIDIA software, which the CEO calls the single most important thing; CUDA-X libraries are being accelerated for AI agents.
- Secure supply and finance the buildout
- Supply and capacity commitments rose from $119 billion to $279 billion in one quarter, and NVIDIA guarantees land, power and shell and co-invests with capital providers for select customers.
Latest results
In Q2 fiscal 2027 (quarter to July 26, 2026) revenue was $96,221 million, up 106% year on year and 18% sequentially, with gross margin of 75.0% and net income of $59,688 million. Data Center revenue reached $89,023 million (+117%) on the Blackwell Ultra ramp, and Edge Computing $7,198 million (+27%). NVIDIA guides Q3 revenue of $108 billion, plus or minus 2%, with no China data center compute revenue assumed, and preliminary fiscal 2028 revenue growth of approximately 70%, described as supply-constrained. Gross margin is reset for higher memory costs to 74% in Q3, bottoming at 71% to 72% in Q4. The quarterly dividend rose from $0.01 to $0.25 per share.
Investments and deals
- Groq licence and team
- December 2025 non-exclusive licence for Groq's language processing unit technology and hiring of certain employees; $13.0 billion paid at closing and $4 billion payable within a year.
- Ecosystem equity investments
- Equity investments of $99 billion and equity commitments of $25 billion, including nearly 50 billion in frontier AI labs.
- SB Energy PORTS campus guarantees
- Guarantees capped at $105 billion for leases at a 4.25-gigawatt campus in Pike County, Ohio, hosting NVIDIA compute under 20-year leases to OpenAI.
- Supply and cloud commitments
- Total future commitments of $366 billion, mainly memory and manufacturing capacity, plus $36 billion of AI cloud agreements with a possible revenue share.
- Capital return and debt
- $19.7 billion of buybacks in Q2 FY27, an $80.0 billion new authorization, and $25.0 billion of senior notes issued in June 2026.
Challenges and risks
- Supply constraints
- NVIDIA says its entire supply chain is challenged and supply is a bottleneck at least through the end of fiscal 2028, with lead times of more than 12 months.
- Memory costs and margin
- Memory prices rose more than expected, so gross margin guidance was reset to a bottom of 71% to 72% in Q4 FY27.
- China export controls
- A $4.5 billion H20 charge in fiscal 2026 and a $0.4 billion H200 charge in fiscal 2027; NVIDIA says it was effectively foreclosed from China's data center compute market.
- Customer concentration and financing risk
- One direct customer was 22% of fiscal 2026 revenue; the 10-Q adds a risk factor on commitments and guarantees, including the risk that OpenAI does not perform its obligations.
- Competition from in-house chips
- AMD, Huawei, Intel and cloud companies with in-house chips (Alphabet, Amazon, Microsoft and others) compete, as do custom chips at AI labs.
What leadership says
“Our single most important thing today is the install base of CUDA”
“Remember, Nvidia is an accelerated computing company, not a GPU company.”
“I think the answer is our entire supply chain is challenged.”
“we know some will call this circular financing. We see it differently.”
Where a supplier can help
- Supply chain capacity
- Tied to the supply constraints challenge and the $279 billion of supply commitments: memory, packaging, rack assembly and component suppliers that add capacity for Vera Rubin.
- Land, power and shell for AI factories
- Tied to the buildout financing priority: power, data center construction, cooling and energy partners, which the CEO wishes were stood up faster.
- Compute infrastructure for R&D
- Tied to R&D spending, where compute and infrastructure costs rose 127% in Q2 FY27: cloud, storage and engineering tooling for 31,000 R&D staff.
- Ecosystem and channel enablement
- Tied to growth beyond hyperscalers: system integrators, neoclouds and software partners that deploy NVIDIA platforms for sovereign and enterprise customers.
- Agentic tooling
- Tied to the CUDA install base priority: software vendors whose tools agents will use, which NVIDIA wants accelerated through CUDA-X libraries.
Tamtam's reading of the priorities and challenges above, not the company's own words.
Selling to companies like NVIDIA?
Tamtam finds the companies that match your ideal customer, researches each account and tells you who to contact and when.
Find similar companiesKey documents
Read in full by Tamtam; each summary links to the document itself.
- Annual Report on Form 10-K for fiscal year 2026 (year ended January 25, 2026)
- NVIDIA describes itself as "a data center scale AI infrastructure company reshaping all industries": a fabless platform company selling GPUs, CPUs (Grace), DPUs, NVLink/InfiniBand/Ethernet networking, rack-scale systems and software (CUDA, CUDA-X, NVIDIA AI Enterprise) to cloud providers, AI model makers, enterprises, the public sector, gamers and carmakers.
- Fiscal 2026 (year to January 25, 2026): revenue $215,938 million (+65%), operating income $130,387 million (+60%), net income $120,067 million (+65%), diluted EPS $4.90 (+67%); gross margin 71.1% (75.0% a year earlier).
- Two reporting segments: Compute & Networking $193,479 million (+67%) and Graphics $22,459 million (+57%). By end market, Data Center was $193,737 million (+68%; compute $162,361 million, networking $31,376 million, networking +142%), Gaming $16,042 million (+41%), Professional Visualization $3,191 million (+70%), Automotive $2,349 million (+39%).
- Blackwell represented the majority of Data Center revenue; Blackwell Ultra (GB300) shipped from Q2 FY26; the Rubin platform is expected to start production shipments in the second half of fiscal 2027, with "up to a 10x reduction in cost per token compared to Blackwell".
- Strong customer concentration: one direct customer was 22% of revenue and another 14%; US-headquartered customers $149,617 million of revenue; customers outside the US 31% (41% a year earlier).
- Ecosystem investing at scale: $17.5 billion invested in private companies and infrastructure funds, $3.5 billion of land, power and shell guarantees, a Groq licence and team hire ($14.4 billion goodwill), $95.2 billion of manufacturing, supply and capacity commitments and $27 billion of multi-year cloud service commitments.
- China: a $4.5 billion H20 charge in Q1 FY26; about $60 million of licensed H20 revenue; the company says it was "effectively foreclosed from competing in China's data center computing/compute market".
- About 42,000 employees in 38 countries (31,000 in R&D); turnover 3.7%. Capital return: 282 million shares repurchased for $40.4 billion; $58.5 billion of authorization remaining.
- Quarterly Report on Form 10-Q for Q2 fiscal 2027 (quarter ended July 26, 2026)
- Q2 FY27 (quarter to July 26, 2026): revenue $96,221 million (+106% year on year, +18% sequentially); gross margin 75.0%; operating income $63,734 million; net income $59,688 million; diluted EPS $2.46.
- New market presentation: Data Center $89,023 million (+117%), split into Hyperscale $48,710 million and AI Clouds, Industrial, & Enterprise (ACIE) $40,313 million (+138%); Edge Computing $7,198 million (+27%).
- Segments: Compute & Networking $88,299 million (+114%), Graphics $7,922 million (+46%). Blackwell Ultra drove growth; Vera Rubin began production shipments in Q3 FY27.
- Supply and capacity commitments rose from $119 billion to $279 billion in one quarter; total future commitments $366 billion, plus $36 billion of AI cloud agreements.
- New financing role: guarantees capped at $105 billion for SB Energy's 4.25-gigawatt PORTS campus, leased to OpenAI; memorandums with capital providers to raise more than $500 billion; equity investments of $99 billion and $25 billion more committed.
- China: a $0.4 billion H200 charge; Hopper shipments to China under 1% of Data Center revenue.
- Capital: $19.7 billion of buybacks in the quarter, $80.0 billion new authorization, quarterly dividend raised from $0.01 to $0.25; $25.0 billion of senior notes issued in June 2026.
- Q2 fiscal 2027 earnings call (Jensen Huang, Colette Kress)
- Revenue of $96 billion, more than doubled year on year; Data Center $89 billion (+18% sequentially): Hyperscale $49 billion (+13%), ACIE (neoclouds, industrial, enterprise) $40 billion (+25% sequentially, +138% year on year).
- Guidance: Q3 revenue of $108 billion, plus or minus 2%, with no China data center compute revenue assumed; preliminary fiscal 2028 revenue growth of approximately 70%, which Kress called "a supply-constrained outlook". Customers' forecasts point to growth doubling.
- Margins reset for memory costs: 74% in Q3, a bottom of 71% to 72% in Q4, then 72% to 73% in fiscal 2028 as price increases take effect.
- Revenue per gigawatt rises from about $18 billion (Hopper) to $25 billion (Blackwell) to $40 billion (Vera Rubin). Vera Rubin began production shipments and should be about 20% of Data Center revenue in Q3.
- New lines: Grace CPU revenue above $5 billion over twelve months, with CPU revenue expected to more than double in fiscal 2028; Groq 3 LPX racks in full production; AWS to deploy an additional 2 million GPUs.
- Financing role: nearly 50 billion invested in frontier AI labs; six capital providers to raise over $500 billion; a revenue-share model with neoclouds ("We're not making loans."). AI labs backed by the balance sheet should be roughly a quarter of the business next year.
- Huang: the whole supply chain is constrained ("The unconstrained would be a lot higher."); non-hyperscale customers are about half the business and growing 100% a year.
- Lex Fridman Podcast #494: Jensen Huang, NVIDIA - The $4 Trillion Company & the AI Revolution
- Extreme co-design: the problem "no longer fits inside one computer", so NVIDIA optimizes GPU, CPU, memory, networking, switching, power and cooling together; Huang has a direct staff of 60 and holds no one-on-ones.
- Moat: "Our single most important thing today is the install base of CUDA"; second is an ecosystem that puts one architecture into every cloud, OEM, car, robot and satellite. He cites putting CUDA on GeForce, which raised GPU cost 50% when NVIDIA was a 35% gross margin company.
- Scaling: four scaling laws (pre-training, post-training, test time, agentic), with intelligence scaling "by one thing, and that's compute". Moore's Law would have given about 100 times in 10 years; NVIDIA scaled computing a million times.
- Supply chain: each Vera Rubin rack has 1.3 million to one and a half million components from 200 suppliers; NVIDIA may produce about 200 Vera Rubin pods a week. He says he does not lose sleep over supply because suppliers have committed what he asked for.
- Power: grids run at about 60% of peak most of the time; he wants data centers that degrade gracefully and utilities that sell interruptible power.
- Growth: he sees nothing preventing NVIDIA from becoming a "$3 trillion revenue company" and expects $1000 per million tokens for premium intelligence; agents, OpenClaw especially, are "the iPhone of tokens".
- Jobs and AGI: "I think we've achieved AGI" (for a short-lived billion-dollar app); he expects more software engineers at NVIDIA, and coders to rise from 30 million to about 1 billion.
- An Interview with Nvidia CEO Jensen Huang About Accelerated Computing (Stratechery, after the GTC 2026 keynote)
- Recorded right after the GTC 2026 keynote, which retold the CUDA-X story because AI agents will use human tools (Excel, Photoshop, Synopsys and Cadence tools, SQL) that must be "super-accelerated".
- AI factories: a gigawatt factory costs "probably $50, $60 billion", of which about $15 to 17 billion is land, power and shell; NVIDIA must give customers the confidence to commit that much.
- CPUs: NVIDIA was "never against CPUs"; agentic tool use needs the fastest single-threaded CPU, and Vera has three times the bandwidth per core of any CPU ever designed.
- Groq: NVIDIA licensed the technology and hired the team (not its cloud business) to extend inference to very low latency, splitting decode work with Vera Rubin; worth adding mainly for Anthropic-like or OpenAI-like businesses with high-value coding tokens.
- Supply: "I think it's probably close on everything. You couldn't double anything, really." His main wish is that land, power and shell were stood up faster.
- China and policy: an American tech stack is needed in China; Huawei had a record year; AI is "a five-layer cake" and the U.S. must win every layer; he criticizes "doomers" in Washington.
- NVIDIA is "an accelerated computing company, not a GPU company": it builds vertically, opens horizontally, and does not intend to become a solutions or service provider.
Questions about NVIDIA
What does NVIDIA do?
NVIDIA designs GPUs, CPUs, networking, rack-scale systems and software such as CUDA for AI data centers, gaming, professional visualization and autonomous vehicles; it is fabless, with TSMC and Samsung making its wafers.
What is NVIDIA's revenue?
Fiscal 2026 revenue was $215,938 million, up 65%. Q2 fiscal 2027 revenue was $96,221 million, and Q3 is guided at $108 billion, plus or minus 2%.
How many employees does NVIDIA have?
NVIDIA has about 42,000 employees in 38 countries, 31,000 of them in R&D.
Who leads NVIDIA?
Jensen Huang, co-founder, has been President and CEO since 1993; Colette M. Kress is EVP and CFO.
What is NVIDIA's strategy?
NVIDIA pursues accelerated computing through extreme co-design of chips, networking, systems and software, selling full AI factories on a one-year product cadence and extending the CUDA ecosystem beyond hyperscalers.
Written by Tamtam from the company's own published documents listed above, read in full. Figures are as those documents state them, for the periods they name. (2026-11-17)